The Workforce Capacity Planning Model for Operations

Workforce capacity planning answers one question: how many people does this operation need to do next year’s work? This model answers it from the bottom up. You list the activities your team performs, record the measured hours and volume behind each one, score each activity for automation potential, and run five steps: measured hours, automatable hours, freed capacity, FTE equivalent, and staffing scenarios. What comes out is a headcount range you can defend line by line, before anyone commits to a number.

A fill-in capacity model for the moment before a headcount number is agreed.

The question reaches you in one of three forms. Volume is growing and the budget asks whether you need to hire. The AI team says automation will free up capacity and asks how much. Or the board has already picked a number and wants to know if the operation can live with it. All three are the same calculation, and most operations run it on the org chart and a planning estimate.

One principle runs through the model: headcount is an output, never an input. You do not start from the number of people. You start from the work, and the number of people falls out at the end.

The Five Inputs the Model Needs

The model runs on five inputs: the activities the team performs, the measured hours each activity consumes, the volume that drives each one, an automation score per activity, and a short list of planning assumptions. The first four are recorded per activity, one row each. The fifth is set once for the whole team.

Two of these are routinely estimated rather than measured: the hours per activity and the automation score. They are also the two inputs that move the answer most, which is why the worked example below tests both.

The Formula Chain, From Measured Hours to Headcount

The workforce capacity planning formula here is a chain of five steps, and each one feeds the next. Work it per activity row first, then total the rows. Doing the arithmetic row by row is what lets you see where the capacity comes from, instead of trusting one blended percentage.

1. Measured hours. Record team hours per week for each activity. Divide by volume to get minutes per item, which you need for growth and as a sanity check: minutes per item = hours × 60 ÷ volume.
2. Automatable hours. Multiply each activity’s hours by its automation score: automatable hours = measured hours × automation score.
3. Freed capacity. Not every automatable hour comes back. Ramp time, exceptions that stay manual, and review of the automated output all take a share. Multiply by a realization factor: freed hours = automatable hours × realization factor. Our Mandate Response Framework applies a 20 to 30% discount at this point, which is a realization factor of 70 to 80%.
4. FTE equivalent. Convert hours into people using productive hours, meaning contracted hours minus shrinkage (paid time away: vacation, holidays, sick time, training): productive hours per FTE = contracted hours × (1 – shrinkage), then FTE equivalent = freed hours ÷ productive hours per FTE.
5. Staffing scenarios. Grow the volume-driven rows by next year’s forecast, subtract the freed hours, and divide by productive hours: FTE needed = (next year’s hours – freed hours) ÷ productive hours per FTE. Run it three times (no automation, first-phase automation, full plan) and compare each result with today’s headcount and the number of people who will leave anyway.

Scoring automation potential

Score the activity, never the role. A role is a bundle of activities, and a claims processor or loan processor spends part of every day on work at each end of this scale.

Quarter steps are deliberate. An automation score of 43% implies a precision nobody has at the planning stage, and a coarse scale keeps the argument on the activity rather than on the decimal.

The Workforce Capacity Planning Template

The template has two parts: an activity sheet with one row per activity, and an assumptions block you fill in once. Copy both into a spreadsheet or a document. Leave the automatable hours column blank until every activity has measured hours, because scoring before measuring is how the rows with the most visible work get the most attention.

Activity sheet

Assumptions

Rebuilding the model in a spreadsheet

To let a spreadsheet run the chain, enter up to 10 activities in rows 2 to 11 and the assumptions in column M. The formulas below carry the whole chain, including next year’s growth, so changing one assumption reruns every scenario.

ASSUMPTIONS (column M)
M2  Current FTE
M3  Contracted hours per FTE per week
M4  Shrinkage, as a decimal
M5  =M3*(1-M4)                       Productive hours per FTE
M6  Realization factor, as a decimal
M7  Volume growth next 12 months, as a decimal
M8  Annual attrition rate, as a decimal

ACTIVITY ROWS (row 2 shown, copy down to row 11)
A2  Activity name
B2  V for volume-driven, F for fixed
C2  Volume per week
D2  Measured hours per week
E2  =IF(C2>0,D2*60/C2,"")             Minutes per item
F2  Automation score: 0, 0.25, 0.5, 0.75 or 1
G2  =D2*F2                            Automatable hours today
H2  =D2*(1+IF(B2="V",$M$7,0))         Hours next year before automation
I2  =H2*F2*$M$6                       Freed hours next year
J2  =H2-I2                            Hours needed next year

TOTALS (row 12) AND OUTPUTS
D12 =SUM(D2:D11)   G12 =SUM(G2:G11)   H12 =SUM(H2:H11)
I12 =SUM(I2:I11)   J12 =SUM(J2:J11)
M10 =M2*M5-D12                        Reconciliation gap, hours per week
M11 =G12*M6/M5                        FTE freed at today's volume
M12 =H12/M5                           FTE needed next year, no automation
M13 =J12/M5                           FTE needed next year, full plan
M14 =M2*M8                            Expected leavers in 12 months
M15 =M13-M2                           Change against today, full plan

For a first-phase scenario, copy the sheet and set every automation score to 0 except the activities you would automate first.

Two checks before you trust the output

The reconciliation check: Measured hours across all rows should add up to roughly current FTE times productive hours (cell M10 above). If the rows come up well short, there is work on the team’s calendar that is not on your activity list, and it is usually the unofficial work. If they come up well over, some hours are counted twice.

The minutes check: Read the minutes per item column. If an activity takes three minutes on paper and your team leads laugh when they see it, the hours or the volume behind that row are wrong, and every step after it inherits the error.

A Worked Example: One Consumer Lending Operations Team

Here is the model filled in for one hypothetical operation: a consumer lending operations team of 50 FTE at a regional bank, processing about 900 loan files a week. The numbers are illustrative, not client data. They exist to show how the chain behaves, including how far the answer moves when one input changes.

Assumptions. 50 FTE on 40 contracted hours, 15% shrinkage, so 34 productive hours per FTE and 1,700 productive hours a week across the team. Realization factor 75%. The business plan expects 10% more volume next year, and annual attrition runs at 12%.

The rows reconcile: 1,700 measured hours against 1,700 productive hours. Automatable hours come to 600, or 35% of the team’s time.

The chain at today’s volume. 600 automatable hours at a 75% realization factor is 450 freed hours a week. Divided by 34 productive hours, that is 13.2 FTE of freed capacity.

The chain at next year’s volume. Growth lands only on the volume-driven rows (1,555 of the 1,700 hours), adding 155.5 hours a week. Next year’s work before automation is 1,855.5 hours, which needs 54.6 FTE. Automatable hours grow with the volume to 660, of which 495 come back at 75% realization. That leaves 1,360.5 hours, which needs 40.0 FTE.

How sensitive the answer is. Change the realization factor and scenario C moves by nearly three FTE in either direction: 42.9 FTE at 60%, 40.0 at 75%, 37.1 at 90%. Change one automation score, intake at 25% instead of 50%, and it moves from 40.0 to 42.2. Neither change is visible from the org chart, and both come from the two inputs that planning estimates usually supply.

The honest answer for this team is a range, from about 37 to 55 FTE, and the size of that range tells you how much of the decision still rests on inputs that nobody has measured.

From FTE Numbers to Staffing Scenarios

A capacity model does not make the headcount decision. It prices the options, so the decision is made knowingly. Bring the three scenarios together, each tied to the activity sheet, and show the range rather than a single figure.

Three habits keep the scenarios credible:

• Set the FTE change against attrition. In the example, scenario C needs ten fewer roles and six people leave in a normal year anyway. That is a staffing plan built on not backfilling, which is a very different conversation from a reduction.
• Phase the automation. Scenario B is the first phase of scenario C. It absorbs growth without a single new hire, and it proves the realization factor on real work before anyone plans on the full number.
• Keep the protected work visible. The 0% rows (exceptions, condition clearing, coordination) do not shrink in any scenario. Showing them explicitly is what makes the rest of the sheet believable.

If the scenarios are going to a board or a CFO, the Mandate Response Framework sets out how to present risk-adjusted options, and justifying operations headcount to the CFO covers the finance side of the same conversation. For a banking operation specifically, the AI headcount mandate in banking operations covers the regulatory constraints that decide which rows can move.

Where the Inputs Usually Go Wrong

Most capacity models fail on their inputs rather than their arithmetic. Three inputs cause most of the trouble, and each one pushes the answer in a predictable direction.

Hours per activity come from estimates. People are unreliable reporters of their own time, even of total hours. A study in the Bureau of Labor Statistics’ Monthly Labor Review compared estimated and diary-recorded work time and found that employed respondents overestimate their work hours by 5 to 10% against the hours in their own time diaries, and that “most of the overestimation is accounted for by respondents who estimate longer work hours.” Splitting a week across eight activities is a harder recall task than totaling it, so an estimated activity sheet can be expected to carry at least that much error, concentrated in exactly the rows people feel busiest in.

Automation scores get assigned to roles. A score given to “loan processor” averages intake work at 50% with exception work at 0% and produces a number that describes no actual activity. Score the rows, then let the roles follow. How to know what to automate in operations goes further on the selection criteria.

The activity list leaves out unofficial work. Status requests, re-keying, and chasing documents rarely appear in process maps. In the example, status requests and re-keying alone take 290 hours a week, 17% of the team’s time. Leave them out and the reconciliation check fails, or worse, the missing hours get spread across the official rows and inflate their minutes per item.

This is also where the model meets a limit of any spreadsheet. The sheet can hold measured inputs, but it cannot measure them. For the question of which software should hold the model once it is built, capacity planning software for operations compares six kinds of tools and what each one plans against. For why the unit of the answer should be hours rather than roles, see how many roles AI can actually replace in back office operations.

The Model in One Look

Frequently Asked Questions

What is workforce capacity planning? Workforce capacity planning is the practice of matching the hours your people have available against the hours the work will require, so you can decide whether to hire, hold, or resize before a gap opens. This model does it from the bottom up, activity by activity, and treats automation as one of the inputs rather than a separate project.

What is the formula for workforce capacity? Available capacity per person is contracted hours minus shrinkage, the time away for vacation, holidays, sickness, and training. The FTE you need is the hours the work requires divided by that figure. This model adds automation to the chain: measured hours times automation score times realization factor gives freed hours, which come off next year’s requirement before you divide.

Is there a workforce capacity planning template in Excel? Yes. The template on this page rebuilds in any spreadsheet: one row per activity, an assumptions block, and the cell formulas listed above. It is free and needs no email address.

How do you include AI in a workforce capacity plan? Score each activity for how much of its time software could realistically take on, apply a realization factor for ramp time and the exceptions that stay manual, and subtract the result from next year’s hours. Do it per activity, not per role, and present the answer as a range, because the automation score is the input most likely to be wrong.

What realization factor should I use? Start between 70% and 80% of automatable hours, which matches the discount in our Mandate Response Framework. Use the low end where exception rates are high or the work is regulated, and replace the assumption with the rate you actually measure after the first phase.

How is a capacity model different from capacity planning software? The model is the arithmetic; software is where the arithmetic lives and gets refreshed. Most capacity planning software takes effort per item and hours as inputs rather than measuring them, so the model’s inputs matter whichever tool you use.

Where do the measured hours come from? From activity data captured across the team over several weeks, classified at the task level rather than the application level. Manager estimates and surveys can fill a first draft of the sheet, but they should not carry a headcount decision.

The model runs on two inputs most operations do not have: measured hours per activity and an automation score for each one. The Summit Trails 90-day Phase 1 assessment produces both: hours captured and classified at the activity level, each activity scored for automation potential, and a deliverable that includes a staffing model analysis. Book a 30-minute strategy call, bring your filled-in sheet, and we will walk through which rows would move once they are measured.

To see how activity is captured and classified, read our approach, and the results page shows what the output looks like.

Use this model with: the Mandate Response Framework, the AI Readiness Scorecard to check whether your data can support the inputs, and the other free tools for operations leaders.

Rather talk it through? Wendy runs a 30-minute live session on answering a headcount question with evidence, with upcoming dates on the webinar page.